Current risk stratification in acute myeloid leukemia (AML) is based on clinical variables such as age and traditional cytogenetics. Recently, molecular genetic analysis has proven to be useful in risk stratification and therapeutic management of AML patients. For example, FLT3, NPM1, and CEBPA mutational analysis was found to improve risk stratification in patients without karyotypic abnormalities. However, all current prognostication schema fail to robustly separate patients who will be cured by chemotherapy from those who will relapse. To address this problem, we have performed studies to enhance molecular prognostication using both genetic and epigenetic markers and, importantly, to make these assays clinically feasible. The Levine Laboratory recently completed mutational analysis of 16 genes mutated in AML in a large cohort of patients with de novo AML and identified novel predictors of outcome, including TET2, PHF6, DNMT3A, and IDH2 R140Q mutations. Importantly, combinations of these genetic variables provide enhanced prognostic ability for many AML patients, leading to predictions of survival of greater than 80% in these sub-groups. In other cases, combinations of genetic variables provide only modest discrimination of prognosis. Additionally, the Melnick laboratory has recently demonstrated through genome-wide DNA methylation analysis that response to therapy for AML patients can be predicted with a fifteen-gene DNA methylation panel. This panel was identified using a genome-wide methylation profiling platform not currently amenable to use in the clinical setting. Taken together, these studies lead to the hypothesis that AML is a disease characterized by both genetic and epigenetic changes and that optimal prognostication schema will need to incorporate both types of markers. To make the prognostic use of DNA methylation profiling feasible in clinical practice, we have developed a microsphere-based methylation profiling (MELP) assay which uses technology widely available in molecular pathology laboratories (see below for details). Initial evaluation demonstrates that the assay robustly reproduces results generated with the original HELP (HpaII tiny fragment enrichment by ligation-mediated PCR) research-based methylation platforms and predicts outcome in a previously described AML cohort (HOVON). In addition, we have implemented a next generation based DNA sequencing methodology to determine genetic mutations in AML and applied this methodology to a recently annotated UPENN AML group (described below). In this proposal, we will further develop a robust, clinically feasible, AML prognostication system combining genetic, epigenetic and clinical features. We will initially focus on confirming a robust DNA methylation based clinical assay and then integrate this assay into a multivariate risk assessment system being developed. The latter system should serve as a platform for ongoing development of a fully integrated prognostic system as we move towards realization of the goal of personalized cancer diagnostics.